It has long been a goal of device physicists to create an optical computing and switching technology to overcome the limitations of conventional electronic circuits such as capacitive charging, electrical cross-talk, lossy transmission lines, and dispersion. The low cross-talk, large signal propagation velocity and the wide tolerances of optical drivers make an optical switching technology particularly attractive for communication over distances larger than .apprxeq.0.5 cm. Thus, high speed optical switches are proving ever more important for technologies dominated by electronics.
The requirements which must be fulfilled by an optical or optoelectronic switching technology include: 1) it must be capable of planar integration; ii) it must operate at .gtoreq.0.5 Ghz; iii) the device sizes must be relatively small (&lt;20 .mu.m.times.100 .mu.m); iv) fabrication should be simple, avoiding sophisticated hybridization of optics and microelectronics; v) isolation of signal sources from the outputs signals of their loads; vi) devices must amplify the signal; vii) signal fan out must be greater than unity; viii) the technology must include the capability of latching and/or delaying signals. Features which are very advantageous in an optical or optoelectronic switching technology include: 1) simple signal interfacing with conventional microelectronic circuits; ii) the flexibility of being able to generate and detect multiple wavelengths of light; iii) the flexibility of being able to detect light at one wavelength and produce one or two separate and independent output wavelengths.
One approach to an optoelectronic switching technology lies in light-emitting diodes with multiple pn junctions. A variety of bistable pnpn light-emitting diode structures have been reported in the literature. These structures include homojunction thyristors constructed from GaAs, AlGaAs, GaP, AlGaAs and GaAs.sub.1-a P.sub.x, as well as heterojunction thyristors constructed from GaAs/AlGaAs and InGaAsP/InP. In addition to pnpn thyristors, bistable light-emitting pnin, pnpnp and pnpnpn diodes have been demonstrated.
Light-emitting pnpn diodes have been fabricated which produce light primarily at the center pn junction or at only one of the two outer pn junctions. In an article by C. P. Lee, A. Gover, S. Margalit, I. Samid and A. Yariv, entitled "Barrier-Controlled Low-threshold PNPN GaAs Heterostructure Laser," Appl. Phys. Lett., vol. 30 (1977), p. 535, there is disclosed a structure which is similar to the present structure. It is a heterostructure laser with two active regions at the outer pn junctions. However, the structure separates the two outer pn junctions with thick (.gtoreq.2.1 .mu.m) AlGaAs layers with uniform composition such that the base widths are large relative to the diffusion length. This results in relatively weak coupling between the light producing outer pn junctions. Also, the structure does not use a single optical cavity/waveguide for both light producing junctions, but rather uses two relatively weakly coupled waveguides. If the two laser regions do not share a single optical cavity, it becomes much more difficult to interface and mode match the cavity with optical waveguides or fibers.
Optical signals can be guided in a dielectric waveguide in which one or more planar dielectric layers, with refractive indices given by n.sub.2, . . . n.sub.L-1, is sandwiched between two outer dielectric layers with refractive indices given by n.sub.1 and n.sub.L. In order for a guided optical mode to exist, the refractive indices (n.sub.1, n.sub.L) of the two outer layers must be smaller than the refractive index (n.sub.2 -n.sub.L-1) of at least one of the sandwiched layers.
A heterostructure thyrister which employs a single optical cavity in the directions perpendicular to the direction of the propagation of stimulated emission is desirable in order to allow simple optical coupling with external optics (including lenses, optical fibers and integrated waveguides). When light travels along two parallel waveguides, it becomes difficult to match the optical modes at the ends of the coupled waveguides with simple optics or a single external waveguide. However, when different dielectric cavities are oriented along the direction of propagation of the optical mode, efficient coupling of external optics to and from the dielectric cavities is relatively simple.
Thus, it is necessary to distinguish between a single dielectric waveguide and coupled dielectric waveguides, when the desired optical modes propagate parallel to the planes which couple different dielectric layers. Since the mirror losses of semiconductor lasers generally favor transverse electric optical modes over transverse magnetic optical modes, this distinction can be made in terms of the lowest order transverse electric mode, which will be referred to as the fundamental mode.
Each guided transverse electric mode is characterized by a specific propagation constant B, whose value satisfies the condition: EQU MAX{n.sub.1.sup.2,n.sub.L.sup.2 }k.sub.0.sup.2 &lt;B.sup.2 &lt;MAX{N.sub.2.sup.2, . . . n.sub.L-1.sup.2 }k.sub.o.sup.2, (1)
where MAX{n.sub.1.sup.2, n.sub.L.sup.2 } denotes the square of the refractive index of the outer dielectric layer with the larger refractive index. Similarly, MAX{N.sub.2.sup.2, . . . n.sub.L-1.sup.2 } denotes the square of the refractive index of the inner, sandwiched dielectric layer with the largest refractive index. B is the propagation constant of the mode and k.sub.o is the free-space propagation constant of light with the desired energy. The lowest order transverse electric mode has the largest propagation constant and propagation velocity. This lowest order transverse electric mode always exists for a planar dielectric waveguide capable of guiding an optical mode.
The calculations to distinguish between a single dielectric waveguide and multiple coupled dielectric waveguides are made in terms of a stack of dielectric layers oriented parallel to the y-z plane (using Cartesian coordinates) and optical modes propagating in the positive z direction. The dielectric layers are considered homogeneous and infinite in extent along the positive and negative y directions. The upper and lower of the two outer dielectric layers are considered infinite in extent along the positive and negative K directions, respectively (perpendicular to the plane of the dielectric layers) .
Following the method presented in a standard text [H. C. Casey, Jr. and M. B. Panish, Heterostructure Lasers: Part A, (Academic Press: New York, 1978)], the optical electric field of a transverse electric mode has the form: EQU E(x,z,t)=Ey(x,z,t)=Ey(x)Re{exp[j(2.pi.ft-Bz) ]}, (2)
where B is the propagation constant of the transverse electric mode, f is the frequency of the light, t is time and is the square root of negative one. The expression Re{. . . } indicates that the real part of the enclosed expression is to be used. Similarly, the component of the magnetic field along the direction of propagation of the mode (+z direction) has the form: EQU Hz(x,z,t)=Re{(j/2.pi.f.mu.).delta.Ey(x,z,t)/.delta.x}, (3)
where .delta.Ey(x,z,t)/.delta.x is the partial derivative of the component of the electric field in the y direction with respect to position along the x direction.
Precise values for the transverse electric modes are obtained by requiring the magnitude of the optical electric field E(x,z,t) to approach zero as the position in the positive and negative x directions becomes infinitely large, and by also requiring Ey(x,z,t) and Hz(x,z,t) to be continuous in the x direction.
If the square of the propagation constant B is smaller than the product of the square of the free-space propagation constant k.sub.o and the square of the refractive index n(x.sub.1) (in other words, if O&lt;n(x).sup.2 k.sub.o.sup.2 -B.sup.2 =K.sub.p (x).sup.2), solutions for Ey(x) have the form: EQU Ey(x)=Fsin[K.sub.p x]+Gcos[K.sub.p (x)x], (4)
where K.sub.p (x) is a positive real constant which will be referred to as the lateral mode constant. F and G are real constants. Regions in which Ey(x) has the form given by equation (4) will be referred to as non-evanescent regions of the dielectric waveguide.
If the square of the propagation constant B is larger than the product of the square of the free-space propagation constant k.sub.o and the square of the refractive index n(x) (in other words, if O&lt;B.sup.2 -n(x).sup.2 k.sub.o.sup.2 =K.sub.d (x).sup.2), solutions for Ey(x) have the form: EQU Ey(x)=Cexp[K.sub.d x]+Dexp[-K.sub.d (x)x], (5)
where K.sub.d (x) is a positive real constant which will be referred to as the mode lateral decay constant. C and D are real constants. Regions in which Ey(x) has the form given by equation (5) will be referred to as evanescent regions of the dielectric waveguide. K.sub.do (x) is the positive square root of the difference between the square of the propagation constant of the lowest order transverse electric mode B.sub.o and the product of the square of the refractive index of a dielectric layer and the square of the free-space dielectric constant. K.sub.do (x) will be referred to as the fundamental mode lateral decay constant.
The transition point between a single dielectric waveguide and coupled dielectric waveguides can be defined in terms of the absolute value of the integral of the fundamental mode lateral decay constant in the direction perpendicular to the plane of the dielectric layers (x direction) across an evanescent region sandwiched between two non-evanescent regions. Thus, if the absolute value of the integral of the fundamental mode lateral decay constant in the direction perpendicular to the plane of the dielectric layers is larger than 2[ln(200)] across any evanescent region which is sandwiched between two non-evanescent regions, the dielectric structure constitutes two or more coupled dielectric waveguides. (2"[ln200]" is two times the natural logarithm of two hundred.) If the absolute value of the integral of the fundamental mode lateral decay constant in the direction perpendicular to the plane of the dielectric layers is smaller than 2[ln(200)] across all evanescent regions which are sandwiched between two non-evanescent regions, the dielectric structure constitutes a single dielectric waveguide. This condition can be expressed as the following equation: EQU .vertline..intg.K.sub.do (x)dx.vertline.=2[ln(200)].about.10.5966,(6)
where the path of the integral is perpendicular to the propagation direction of the desired optical modes and is the shortest path joining the two abutting non-evanescent regions.
This value is chosen so that {exp[-([.vertline..intg.K.sub.do (x)dx.vertline.)/2]}/2=0.01, corresponding to a factor of 100 attenuation of the optical electric fields in the evanescent region. This degree of attenuation permits the two abutting non-evanescent regions to form distinct optical waveguides, which are coupled by the evanescent optical electric fields in the evanescent region.
Circuits which have been discussed or proposed in the literature which employ light emitting pnpn diodes include simple capacitive relaxation oscillators, a simple trigger flip-flop, optical memories, optical write - electrical read memories, light emitting displays, devices for transforming a spontaneous optical emission into a coherent light pulse, converters for converting dc current into high frequency optical light pulses, repeaters and amplifiers in digital or pulse coded optical communications, high-efficiency, fast-response photodetectors with a controlled spectral sensitivity range, interfaces between optical fibers and digital logic circuits, and for fabricating simple logical AND and OR gates by carefully matching the inputs and the optical switching threshold.